salau-dev/checkam-ng

Open-source, explainable, self-hostable fraud-message risk scoring for Nigeria and West Africa # CheckAm-NG **An open-source, explainable, self-hostable tool that scores text messages for fraud risk — built for the scam patterns hitting Nigeria and West Africa.** > **Scope statement:** We score text messages for fraud risk. We are not a full security platform — yet. --- ## Why this exists Nigeria recorded 1.6 million online cyberattack attempts in the first half of 2026 alone, and its financial sector faces attack volumes far above the global average. Most of the tooling built to fight this targets banks and enterprises. Almost nothing targets the individual reading a suspicious SMS, WhatsApp message, or "bank alert" at 11pm — and nothing in that space is open-source or self-hostable. CheckAm-NG is built to close that specific gap: - **Explainable** — every risk score traces back to named, documented rules. No black-box "trust us" scoring. - **Self-hostable** — run it on your own infrastructure. No mandatory API keys, no forced hosted service, no vendor lock-in. - **Regionally grounded** — detection patterns are built around West African fraud modalities specifically: BVN phishing, SIM-swap social engineering, mobile money/POS fraud, romance-to-crypto scams, and more — not adapted from a US/EU/India scam taxonomy. --- ## How it works Send CheckAm-NG a message. It runs the text through a rule engine, extracts relevant entities (phone numbers, URLs, amounts, crypto addresses), and returns a risk score with a full explanation of exactly which patterns fired and why. ### Example **Request** ```json POST /api/v1/scan { "content": "Your BVN has been flagged for suspension. Call 0803XXXXXXX immediately to verify or your account will be blocked in 24hrs.", "channel": "sms" } ``` **Response** ```json { "ok": true, "data": { "risk_score": 87, "risk_level": "high", "fraud_type": "bvn_phishing", "explanation": "This message impersonates a bank compliance notice and uses urgency combined with a request to call an unverified number about BVN status — a known i …